The continued rise of transcatheter aortic valve replacement (TAVR) as a go-to treatment option for aortic stenosis (AS) has been one cardiology’s biggest trends for several years now. Long-term patient outcomes continue to impress, and other transcatheter therapies are even starting to gain significant momentum.
Can artificial intelligence (AI) help TAVR outcomes get even better? A new analysis written by two cardiologists, and published in full in Diagnostics, explored that very question.[1]
“Within the realm of cardiology, AI plays a significant role by assisting in the interpretation of medical imaging, such as echocardiograms and cardiac MRIs,” wrote first author Mina M. Benjamin, MD, a cardiologist with Saint Louis University Hospital, and co-author Mark G. Rabbat, MD, a cardiologist with Loyola Medicine. “Additionally, AI-powered wearable devices and remote monitoring systems can continuously track cardiac parameters, providing real-time insights and facilitating remote patient management for better overall cardiovascular health. Moreover, AI-driven predictive models can analyze patient data and identify individuals at higher risk of cardiac events, enabling early intervention and better preventive care strategies. The application of AI in TAVR procedures is being explored in order to improve patient selection and procedural planning and optimize patient outcomes and post-implantation valve monitoring.”
4 key ways AI can improve TAVR outcomes
1. Improve the diagnosis of severe aortic stenosis
Diagnosing AS can present certain challenges, the authors wrote, particularly because it may be asymptomatic at first and its symptoms are often linked to other conditions. AI has already shown potential to help clinicians evaluate patients for AS with a number of imaging modalities, including electrocardiography, cardiac CT and echocardiography. AI’s potential to help clinicians during echo exams is especially strong; researchers have found that it can reduce the issues often caused by interobserver variability and overworked echo labs.
“By assisting with echocardiography interpretation, AI can also play a vital role in situations or locations where qualified individuals may not be readily accessible,” the authors wrote. “Deep learning models have demonstrated efficacy in identifying left ventricular hypertrophy, which is associated with AS, and also in accurately measuring left ventricular volumes and dysfunction.”
2. Improve the patient selection process for TAVR procedures
Should a patient with severe AS undergo TAVR? Would surgical aortic valve replacement (SAVR) be a better choice? These are just some of the decisions care teams face on a regular basis, and AI has already shown potential to make those choices more straightforward than ever before. The authors pointed to one AI technique, for instance, that used variables such as a patient’s heart failure data and the presence of peripheral vascular disease to anticipate who should be chosen for TAVR and who may be a better fit for SAVR.[2]
3. Assist with pre-TAVR planning
Benjamin and Rabbat highlighted AI’s ability to help cardiologists with pre-TAVR FFRCT evaluations, noting that advanced algorithms can measure the patient’s aortic valve annular plane, determine the ideal access route and even assist with device sizing.
AI could even help care teams “virtually implant” different device types from different depths, simulating how blood might flow and how the patient’s anatomy may respond in these different scenarios. Many of these algorithms are “not currently ready for prime-time application,” the authors wrote, but they highlight yet another way advanced AI can ensure patients are receiving the best care possible.
4. Predicting a TAVR patient’s mortality risk
“Predicting mortality before TAVR is of paramount importance as it allows clinicians to assess the potential risks and benefits of the procedure for individual patients, aiding in informed decision-making,” the authors wrote. “Accurate mortality prediction helps optimize patient selection, enhance procedural planning, and improve patient outcomes by tailoring treatment strategies based on individual risk profiles.”
Typical cardiovascular risk scores are not necessary as helpful when focusing on TAVR, the authors noted, and AI has helped several research teams develop their own risk scores for this specific purpose. One machine learning-based risk score tracked data from more than 10,000 patients, for instance, and had an area under the ROC curve of 0.92 when it came to predicting in-hospital mortality after TAVR.[3]
In addition, AI models can be developed to anticipate the risk of other specific patient outcomes, including bleeding events or heart failure hospitalizations. Research teams have been working on these algorithms for many years now, helping ensure the care teams of the future will be able to click a single button and suddenly gain access to a wide variety of predictions they can confidently use to make treatment decisions.
A word of caution about AI vs. trained specialists
“While AI holds immense potential for advancing TAVR procedures, challenges must be addressed to ensure its safe and responsible integration,” the authors wrote.
Some of the current challenges clinicians still face when trying to use AI include data privacy, publication bias, selection bias, the use of outdated data and the “unintentional continuance of historical biases/stereotypes in the data that can lead to inaccurate conclusions.”
Benjamin and Rabbat also noted that the innerworkings of many of these AI algorithms are still “only partially understood.”
While these algorithms may go on to be game-changers for the diagnosis and treatment of severe AS, it is still relatively early. They said there is still much more TAVR research that needs to be done before cardiologists and other cardiology professionals can truly know how this technology will impact patient care.
“The advanced artificial intelligence algorithms described in this paper are promising tools that have, and will, further enhance the planning, execution and postoperative follow up of TAVR procedures,” Benjamin and Rabbat concluded. “However, the authors also believe that nuanced clinical judgment by skilled physician teams will remain irreplaceable.”
Michael has more than 16 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.